introducing probabilistic radio propagation models in
TRANSCRIPT
Institut für Telematik
Introducing Probabilistic Radio Propagation Models in OMNeT++ Mobility Framework and
Cross Validation Check with NS-2
A. Kuntz, F. Schmidt-Eisenlohr, O. Graute, H. Hartenstein, M. Zitterbart
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
2
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Why probabilistic models?
Radio propagation modelsDetermine signal strength at receiving nodeSimple deterministic models
Signal strength is a simple function of distance
Consequences of deterministic modelsA radio‘s transmission area is circularAll radios have equal rangeIf I can hear you, you can hear meIf I can hear you at all, I can hear you perfectly
Deterministic models are unrealisticBut standard model for many simulators
[Kotz2004]
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
3
Probabilistic Propagation Models forOMNeT++ Mobility Framework
How about Mobility Framework?
Deterministic Free Space propagation model
All relevant rx events in unit disk graphBasis for MF ChannelControl implementation
Connects nodes in max. propagation distance
LdGGPd rtt
22
2
det )4()(Pr
πλ
=
Pr
d
ChannelControl
Rx
d0
1 maximumpropagation
distance
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
4
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Probabilistic Models
General structureUse deterministic path loss
Average reception powerAdd probabilistic small scale fading
Characterized by some distribution
Newly included modelsLog Normal ShadowingRayleigh RiceNakagami ⎟
⎠⎞
⎜⎝⎛
mdmGammamdNak
)(Pr,~);(Pr det
( )argsdDistrargsdDistr );(Pr~);(Pr detRx
d0
1
Pr
d
[Rappaport1996][Nakagami1960]
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
5
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Integration into Mobility Framework
Only ‚connected‘ nodes receive eventsChannel Control cares about connectionsNeeds access to the RF module
Calculate Pr(d) on each message arrivalRF Models needed in host moduleSNREval keeps track of SNIR
NIC
SNR Evaluator
Decider
MAC
…Application
RF Models
Host
(send direct)
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
6
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Consequences
Rx Probability = {0, 1}Relevant events in max propagation distancecosttx ~ #nodes in ‘circle‘
Rx Probability != 0All events may be relevantAll nodes have to be informedcosttx ~ #nodes in scenario
Pr
d
Rx
d0
1
Deterministic
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
7
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Tradeoff Accuracy vs. Speed
Max propagation distance (MPD) is tradeoff parameterAccuracy vs. speed of simulationLarge MPD slow but accurate simulationSmall MPD fast but inaccurate simulation
Probability of relevant event wrt/ distance
Relevance
Duration of simulation wrt/ MPD
Duration
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
8
Probabilistic Propagation Models forOMNeT++ Mobility Framework
How to determine MPD (sketch)
Track one single eventNode’s goal
receive all relevant eventsRelevant event
Prx(d) > PminFor which distance d holds:
Prob(relevant evt) < accuracyProb(Prx(d) > Pmin) < accuracy
Prx(d) ~ probability distributionCumulative densitiy function
1-CDFPrx(d)(Pmin) < accuracyInvert CDF to get the MPD
Restrict connectivity to distances < MPD
))(()()( minrxmindP PdPProbPCDFrx
≤=
))(()(1 )( minrxmindP PdPProbPCDFrx
>=−
rxP
d
)(dPrxCDF
minPminP
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
9
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Simulation setup
LimitationsProposed general solution to find MPDSolution applied to Nakagami, m=1
General setupTopologies: chain, grid, randomModels: Free Space, NakagamiPlayground
2000m x 2000m, torus enabledUp to 1600 nodes and 100 senders
Fixed MPD for evaluationFree Space 623mNakagami 1000m
Cross validationComparison with NS-2
Performance evaluationSimulation speedMemory usage
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
10
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Cross Validation Check
grid-100, Rx Probability
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
11
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Performance Evaluation Speed
OMNeT++ MF faster than NS-2Nakagami causes more receive events
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
12
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Performance Evaluation Memory
OMNeT++ MF needs less memoryNakagami causes memory increase
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
13
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Summary and Outlook
Introduced Probabilistic Wave Propagation Models to OMNeT++ Mobility Framework
Log Normal Shadowing, Rayleigh, Rice, NakagamiCross validation check with NS-2Performance evaluation
Tradeoff: Accuracy vs. SpeedGeneral solution for max propagation distance (MPD)Concrete solution for Nakagami, m=1
Future WorkMax propagation distance for Nakagami, m>1Analytical results for other modelsImproved modeling for multiple interfering events
AvailabilityExtension online: www.tm.uka.de/sne4omf(Sensor Network Extensions for the OMNeT++ Mobility Framework)
Andreas Kuntz March 7, 2008 Institut für TelematikUniversität Karlsruhe (TH) www.tm.uka.de
14
Probabilistic Propagation Models forOMNeT++ Mobility Framework
Literature
[Kotz2004] Kotz, D.; Newport, C.; Gray, R. S.; Liu, J.; Yuan, Y. and Elliott, C.: Experimental evaluation of wireless simulation assumptions In Proceedings of the 7th ACM international symposium on Modeling, analysis and simulation of wireless and mobile systems (MSWiM04), ACM Press, 2004, 78 – 82[Rappaport1996] Rappaport, T. S.: Wireless Communications: Principles & Practise Prentice Hall PTR, 1996[Nakagami1960] Nakagami, M.: The m-distribution – A general formula for intensity distribution of rapid fadingStatistical Methods in Radio Wave Propagation, Pergamon, 1960